ABOUT ME

I am a Lecturer (equivalent to US Assistant Professor) in the Department of Computing at Imperial College London. My current research interests focus on cloud performance engineering, measurement, and modeling. My Curriculum Vitae includes a list of my research activities and my service record. A quite complete list of my publications is available on DBLP.

Teaching: In 2013/2014 I will be lecturing 436 Performance Analysis and 343 Operations Research.

Group, collaborators, students: I am a member of the Analysis, Engineering, Simulation and Optimization of Performance group (AESOP) in the Department of Computing. The team I work with is currently composed by (in alphabetical order): Karsten Molka, Juan F. Perez, Weikun Wang.

Cloud Computing I recently started investigating the performance and reliability of cloud applications designed with the model-driven engineering paradigm to facilitate multi-cloud portability. My ongoing research in this area is sponsored by the MODAClouds FP7 IP project, see the vision paper:

• MODAClouds: A Model-Driven Approach for the Design and Execution of Applications on Multiple Clouds MISE workshop 2012.

My research in this area has focused on developing techniques to fit empirical datasets and algorithms for statistical inference of metrics that are difficult, or impossible, to measure directly. An example of the latter is the average resource consumption of requests, also called the service demand. In fitting, I focused on a class of hidden Markov models called Markovian Arrival Processes (MAPs). MAPs can fit time series and are easy to embed in a continuous-time Markov chain. This provides a way to descrive in Markov chains events that arrive according to processes that are not Poisson, thus no more restrictions to exponential distributions. A good introduction is Chapter 2 in Alma Riska's thesis.

Relevant Publications:

• Bayesian Service Demand Estimation with Gibbs Sampling IEEE MASCOTS 2013.
• An Offline Demand Estimation Method for Multi-Threaded Applications IEEE MASCOTS 2013.
• Fitting Second-Order Acyclic Marked Markovian Arrival Processes IEEE/IFIP DSN 2013.
• Building Accurate Workload Models using Markovian Arrival Processes   ACM SIGMETRICS 2011, tutorial.
• Trace Data Characterization and Fitting for Markov Modeling Elsevier Performance Evaluation, Feb 2010.
• Estimating Service Resource Consumption From Response Time Measurements  VALUETOOLS 2009.
• KPC-Toolbox: Simple Yet Effective Trace Fitting Using Markovian Arrival Processes  QEST 2008. [code]
• How to Parameterize Models with Bursty Workloads ACM Perf. Eval. Rev., Sep 2008.
• Robust Workload Estimation in Queueing Network Performance Models  Euromicro PDP 2008.

I have published a number of works on methodologies to analyse stochastic models that are common in performance evaluation, mainly queueing network models. Behind the mathematical curtain, these models capture phenomena that we experience daily, for example the sharing of a CPU by multiple processes or the waiting time to download a file. Queueing models represent a sensible way to answer what-if questions about a system experiencing contention and they are a useful decision support tool in many applications of operations research. My research in this area is mainly theoretical and focused on improving the tractability and accuracy of model solution techniques.

Relevant Publications:

• AutoCAT: Automated Product-Form Solution of Stochastic Models MAM7 2012.
• A Class of Tractable Models for Run-Time Performance Evaluation ACM/SPEC ICPE 2012.
• Fluid Analysis of Queueing in Two-Stage Random Environments QEST 2011.
• Approximate Analysis of Blocking Queueing Networks with Temporal Dependence IEEE/IFIP DSN 2011.
• Efficient Parallelization of the Method of Moments for Queueing Networks Using Multi-Modular Algebra VALUETOOLS 2011.
• A Generalized Method of Moments for Closed Queueing Networks  Elsevier Performance Evaluation, Feb 2011.
• Exact Analysis of Performance Models by the Method of Moments  Elsevier Performance Evaluation, Jun 2011.
• CWS: a Model-Driven Scheduling Policy for Correlated Workloads ACM SIGMETRICS, 2010.
• Model-Driven System Capacity Planning Under Workload Burstiness IEEE Transactions on Computers, Jan 2010.
• Approximating Passage Time Distributions in Queueing Models by Bayesian Expansion  Elsevier Performance Evaluation, Nov 2010.
• Product-Form Approximation of Tandem Queues via Matrix-Geometric Methods  NSMC 2010 workshop.
• MAP-AMVA: Approximate Mean Value Analysis of Bursty Systems  IEEE/IFIP DSN 2009.
• The Multi-Branched Method of Moments for Queueing Networks  QEST 2009.
• Geometric Bounds: a Noniterative Analysis Technique for Closed Queueing Networks IEEE Transactions on Computers, Jun 2008.
• A Note on Stable Flow-Equivalent Aggregation in Closed Networks Springer Queueing Systems, Dec 2008.
• Bound Analysis of Closed Queueing Networks with Workload Burstiness  ACM SIGMETRICS 2008. [code]
• Approximate Solution of Multiclass Queueing Networks with Region Constraints  IEEE MASCOTS, Oct 2007.
• An Efficient Algorithm for the Exact Analysis of Multiclass Queueing Networks with Large Population Sizes  ACM SIGMETRICS 2006.
• On Single-Class Load-Dependent Normalizing Constant Equations  QEST 2006.
• A New Class of Non-Iterative Bounds for Closed Queueing Networks  IEEE MASCOTS 2006.
• Bottlenecks Identification in Multiclass Queueing Networks using Convex Polytopes  IEEE MASCOTS 2004. [code]

• If English is your first language, my name is probably difficult to pronounce and spell. Actually, also if you are born in my country (Italy) you may pronunce it wrongly, normally adding a spurious 'gl' in the middle. If you want to learn how to pronunce my name, this website has a reasonable audio recording, although the wide open pronunciation of the 'a' is slighty different from the way it should be. I also keep track of interesting mispellings and errors, you can read the current list here, hopefully you won't contribute to it in the future. :-)

• I am born in the beautiful city of Pavia at the south-west of Lombardy in northern Italy. If you are traveling to Milan or to the lakes, why not making a stop to Pavia as well? It's a renowned rice and wine production area, with a lot of middle age history, a big castle, and the beautiful Ticino river park. Few know that Pavia was the capital of Italy under the Lombard Kingdom, for about 200 years. Since then, Milan has become prominent in the area, but fortunately the good food still belongs to us. :-)